Haskell literals in type inference

Asked Aug 26 '15 at 17:13

Active Aug 26 '15 at 20:34

Viewed 91 times

Let's say I have a function ($10).

Now, since 10 is of type Num a => a, ($10) is surely of type Num a => (a -> b) -> b, and GHCi confirms this.

But when I say let f = ($10), GHC infers f's type to be (Integer -> b) -> b, as if the let-assignment deduced some 'extra' information from the use of the literal.

Does this mean GHC's type inference coerces a concrete type for literals?

But if so, why does :t ($10) return Num a => (a -> b) -> b ?

f :: Num a => (a -> b) -> b
f = ($10)

typechecks, and f (+0.1) yields a correct result, while

g = ($10)

called with g (+0.1) fails, expecting an Integer.

edited Aug 26 '15 at 20:34

Bakuriu

98,325
22
197
231

asked Aug 26 '15 at 17:13

Csongor Kiss

13

Sounds like the monomorphism restriction. Try using `:set -XNoMonomorphismRestriction` in ghci. Btw - you can type `:set -XNoMonom` and then hit TAB to complete it. – ErikR Aug 26 '15 at 17:28
5

Monomorphism + GCCi defaulting rules. – n. m. could be an AI Aug 26 '15 at 17:33
2

That was it! Monomorphism restriction is turned off by default in newer GHC versions, as it happens I was playing around with 7.0.4 – Csongor Kiss Aug 26 '15 at 17:33
2

@CsongorKiss: The monomorphism restriction is turned off by default in newer *GHCis*. It's still on by default in source files. – Antal Spector-Zabusky Aug 27 '15 at 03:18

Haskell literals in type inference

0 Answers0